Method, apparatus, and system for recognizing target object

1. A method for recognizing a target object, comprising: obtaining an image of a target region; determining an image recognition result of target objects in the target region according to the image; obtaining a radio frequency recognition result of the target objects in the target region; and determining a distribution of the target objects in the target region according to the radio frequency recognition result and the image recognition result; wherein the target region comprises at least one sub-region, and the image recognition result comprises N candidate categories having top-N confidence values of the target objects in each of the at least one sub-region and a number of the target objects in each of the at least one sub-region, and determining the distribution of the target objects in the target region according to the radio frequency recognition result and the image recognition result comprises: respectively matching the N candidate categories and the number of the target objects in each of the at least one sub-region in the image recognition result with categories and a number of the target objects of each category in the radio frequency recognition result, according to a descending order of the top-N confidence values and a descending order of the number of the target objects in each of the at least one sub-region in the image recognition result; and obtaining a category and the number of the target objects in each of the at least one sub-region.

2. The method according to claim 1 , wherein determining the image recognition result of the target objects in the target region according to the image comprises: processing the image via a neural network to output the image recognition result of the target objects in the target region.

3. The method according to claim 1 , wherein obtaining the radio frequency recognition result of the target objects in the target region comprises: obtaining the radio frequency recognition result of the target objects in the target region via a radio frequency detection antenna arranged in the target region.

4. The method according to claim 3 , wherein the radio frequency recognition result is obtained by reading a radio frequency recognition tag disposed on each of the target objects via the radio frequency detection antenna.

5. The method according to claim 1 , wherein respectively matching the N candidate categories and the number of the target objects in each of the at least one sub-region in the image recognition result with the categories and the number of the target objects of each category in the radio frequency recognition result according to the descending order of the top-N confidence values and the descending order of the number of the target objects in each of the at least one sub-region in the image recognition result comprises: matching a k-th candidate category of the target objects in an m-th sub-region and a number of the target objects in the m-th sub-region with the categories and the number of the target objects of each category in the radio frequency recognition result; wherein the k-th candidate category is one of the N candidate categories corresponding to a confidence number k; wherein confidence numbers are obtained by ranking the N candidate categories in the m-th sub-region in the image recognition result according to the descending order of the top-N confidence values; wherein the m-th sub-region is a sub-region corresponding to a sub-region number m in the at least one sub-region; and wherein sub-region numbers are obtained by ranking the number of the target objects in each of the at least one sub-region in the image recognition result in a descending order; in response to that the categories in the radio frequency recognition result comprise the k-th candidate category, and the number of the target objects of the k-th candidate category in the radio frequency recognition result is greater than zero, determining that a category of the target objects in the m-th sub-region is the k-th candidate category of the target objects in the m-th sub-region in the image recognition result and the number of the target objects in the m-th sub-region is the number of the target objects in the m-th sub-region in the image recognition result; and deducting the number of the target objects in the m-th sub-region from the number of the target objects of the k-th candidate category in the radio frequency recognition result to obtain an adjusted radio frequency recognition result.

6. The method according to claim 5 , further comprising: in response to that the categories in the radio frequency recognition result do not comprise the k-th candidate category or in response to that the categories in the radio frequency recognition result comprise the k-th candidate category but the number of the target objects of the k-th candidate category in the radio frequency recognition result is not greater than zero, determining that matching is unsuccessful; and determining the m-th sub-region as a sub-region to be matched.

7. The method according to claim 6 , wherein when determining the m-th sub-region as the sub-region to be matched, the method further comprises: in response to that the number of the target objects of each category in the radio frequency recognition result is not greater than zero, determining that one or more candidate categories, in the N candidate categories in the sub-region to be matched, appear in the categories of the radio frequency recognition result, taking the candidate category having a maximum confidence value in one or more appeared candidate categories as the category of the target objects in the sub-region to be matched, and taking the number of the target objects in the sub-region to be matched in the image recognition result as the number of the target objects in the sub-region to be matched.

8. The method according to claim 6 , wherein when determining the m-th sub-region as the sub-region to be matched, the method further comprises: in response to that the number of the target objects of each category in the radio frequency recognition result is not greater than zero, determining that none of the N candidate categories in the sub-region to be matched appears in the categories of the radio frequency recognition result, taking the candidate category having a maximum confidence value as the category of the target objects in the sub-region to be matched, and taking the number of the target objects in the sub-region to be matched in the image recognition result as the number of the target objects in the sub-region to be matched.

9. The method according to claim 6 , wherein when determining the m-th sub-region as the sub-region to be matched, the method further comprises: in response to that the number of the target objects of at least one category in the radio frequency recognition result is greater than zero, taking the category with a first maximum number of the target objects in the radio frequency recognition result as the category of the target objects in the sub-region to be matched with a second maximum number of the target objects in the image recognition result, and taking the number of the target objects in the sub-region to be matched with the second maximum number of the target objects in the image recognition result as the number of the target objects in the sub-region to be matched with the second maximum number of the target objects in the image recognition result; and deducting the number of the target objects in the sub-region to be matched with the second maximum number of the target objects in the image recognition result from the number of the target objects in the category with the first maximum number of the target objects in the radio frequency recognition result to obtain an adjusted radio frequency recognition result.

10. A system for recognizing a target object, comprising: an image collection apparatus, configured to obtain an image of a target region; a radio frequency detection apparatus, configured to obtain a radio frequency recognition result of target objects in the target region; and a processor, configured to determine an image recognition result of the target objects in the target region according to the image, and determine a distribution of the target objects in the target region according to the radio frequency recognition result and the image recognition result; wherein the target region comprises at least one sub-region, and the image recognition result comprises N candidate categories having top-N confidence values of the target objects in each of the at least one sub-region and a number of the target objects in each of the at least one sub-region; and determining the distribution of the target objects in the target region according to the radio frequency recognition result and the image recognition result comprises: respectively matching the N candidate categories and the number of the target objects in each of the at least one sub-region in the image recognition result with categories and a number of the target objects of each category in the radio frequency recognition result according to a descending order of the top-N confidence values and a descending order of the number of the target objects in each of the at least one sub-region in the image recognition result; and obtaining a category and the number of the target objects in each of the at least one sub-region.

11. The system according to claim 10 , wherein respectively matching, by the processor, the N candidate categories and the number of the target objects in each of the at least one sub-region in the image recognition result with the categories and the number of the target objects of each category in the radio frequency recognition result according to the descending order of the top-N confidence values and the descending order of the number of the target objects in each of the at least one sub-region in the image recognition result comprises: matching a k-th candidate category of the target objects in an m-th sub-region and a number of the target objects in the m-th sub-region with the categories and the number of the target objects of each category in the radio frequency recognition result; wherein the k-th candidate category is one of the N candidate categories corresponding to a confidence number k; wherein confidence numbers are obtained by ranking the N candidate categories in the m-th sub-region in the image recognition result according to the descending order of the top-N confidence values; wherein the m-th sub-region is a sub-region corresponding to a sub-region number m in the at least one sub-region; and wherein sub-region numbers are obtained by ranking the number of the target objects in each of the at least one sub-region in the image recognition result in a descending order; in response to that the categories in the radio frequency recognition result comprise the k-th candidate category, and the number of the target objects of the k-th candidate category in the radio frequency recognition result is greater than zero, determining that a category of the target objects in the m-th sub-region is the k-th candidate category of the target objects in the m-th sub-region in the image recognition result and the number of the target objects in the m-th sub-region is the number of the target objects in the m-th sub-region in the image recognition result; and deducting the number of the target objects in the m-th sub-region from the number of the target objects of the k-th candidate category in the radio frequency recognition result to obtain an adjusted radio frequency recognition result.

12. The system according to claim 11 , wherein the processor is further configured to: in response to that the categories in the radio frequency recognition result do not comprise the k-th candidate category or in response to that the categories in the radio frequency recognition result comprise the k-th candidate category but the number of the target objects of the k-th candidate category in the radio frequency recognition result is not greater than zero, determine that matching is unsuccessful; and determine the m-th sub-region as a sub-region to be matched.

13. The system according to claim 12 , wherein when determining the m-th sub-region as the sub-region to be matched, the processor is further configured to: in response to that the number of the target objects of each category in the radio frequency recognition result is not greater than zero, determine that one or more candidate categories, in the N candidate categories in the sub-region to be matched, appear in the categories of the radio frequency recognition result, take the candidate category having a maximum confidence value in one or more appeared candidate categories as the category of the target objects in the sub-region to be matched, and take the number of the target objects in the sub-region to be matched in the image recognition result as the number of the target objects in the sub-region to be matched.

14. The system according to claim 12 , wherein when determining the m-th sub-region as the sub-region to be matched, the processor is further configured to: in response to that the number of the target objects of each category in the radio frequency recognition result is not greater than zero, determine that none of the N candidate categories in the sub-region to be matched appears in the categories of the radio frequency recognition result, take the candidate category a maximum confidence value as the category of the target objects in the sub-region to be matched, and take the number of the target objects in the sub-region to be matched in the image recognition result as the number of the target objects in the sub-region to be matched.

15. The system according to claim 12 , wherein when determining the m-th sub-region as the sub-region to be matched exists, the processor is further configured to: in response to that the number of the target objects of at least one category in the radio frequency recognition result is greater than zero, take the category with a first maximum number of the target objects in the radio frequency recognition result as the category of the target objects in the sub-region to be matched with a second maximum number of the target objects in the image recognition result, and take the number of the target objects in the sub-region to be matched with the second maximum number of the target objects in the image recognition result as the number of the target objects in the sub-region to be matched with the second maximum number of the target objects in the image recognition result; and deduct the number of the target objects in the sub-region to be matched with the second maximum number of the target objects in the image recognition result from the number of the target objects in the category with the first maximum number of the target objects in the radio frequency recognition result to obtain an adjusted radio frequency recognition result.

16. A non-transitory computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is caused to perform operations comprising: obtaining an image of a target region; determining an image recognition result of target objects in the target region according to the image; obtaining a radio frequency recognition result of the target objects in the target region; and determining a distribution of the target objects in the target region according to the radio frequency recognition result and the image recognition result; wherein the target region comprises at least one sub-region, and the image recognition result comprises N candidate categories having top-N confidence values of the target objects in each of the at least one sub-region and a number of the target objects in each of the at least one sub-region; and determining the distribution of the target objects in the target region according to the radio frequency recognition result and the image recognition result comprises: respectively matching the N candidate categories and the number of the target objects in each of the at least one sub-region in the image recognition result with categories and a number of the target objects of each category in the radio frequency recognition result according to a descending order of the top-N confidence values and a descending order of the number of the target objects in each of the at least one sub-region in the image recognition result; and obtaining a category and the number of the target objects in each of the at least one sub-region.